Professional Documents
Culture Documents
North-Holland
247
17 (1985) 247-252
Reserve System,
Received
Washington,
DC 20551,
USA.
6 July 1984
This paper presents a failsafe method for analyzing any linear perfect foresight
computes the reduced-form
solution or indicates why the model has no reduced
model. It describes
form.
a procedure
which either
equation
to t,
whose solution
t > 0.
i
HiX,+, =O,
;= __7
x, = x;,
t=
0)
and X, is an L dimensional
-7,
..)
conditions
(2)
-1
* The views expressed here are those of the authors and do not necessarily represent
We would like to thank Richard Porter for his helpful comments and Anil Kashyap
0165.1765/85/$3.30
B.V. (North-Holland)
Now, suppose
two restrictions:
that the L
A* = 0
then
L real coefficient
{ H,:i = -7,
matrices
. . . , 0) satisfy
the following
X* =O.
{ X,: i = - 7,
= 0.
. , -1).
solution
sequence
X,=
2 B,X,+,,
t > 0.
(3)
,= -7
[H_,...
the procedure
B=
[B_,...
for determining
whether
He],
computes
(a) and
(4)
the reduced-form
B_,]
coefficient
matrix
(5)
3. Summary
of the procedure
The process of casting the original model into the state-space representation
generates a set of
auxiliary initial conditions
that must be satisfied by the state vector in addition
to the stability
conditions and the historical initial conditions. These auxiliary initial conditions in conjunction
with
the stability conditions
determine the models reduced form expressing current values of the model
solution entirely in terms of lagged values. The reduced form together with the historical initial
conditions determine a unique solution to the structural equations converging to the steady state.
We summarize the procedure as a fourteen-step
algorithm. Anderson and Moore (1983) contains
our proof that the procedure verifies assumptions (a) and (b) and computes B after a finite number of
iterations.
Initiulization
(i) Verify that Cf_ _,H,
(2) CH, is non-singular.
Hz= [H_,...
Auxiliary
HO],
Q := null matrix.
initial conditions
(3) Compute the singular values, {p,: i = 1, . . . , L}, and left singular vectors, V, of H,. Sort the
p, small-to-large
and order the columns of V conformably.
If I*, f 0, i = 1, . . . , L, then H, is
non-singular;
go to step (9).
(4) H, is singular. Premultiply the coefficient matrix by VT to annihilate I-rank (H,) rows of HO,
Hz= V*H.
(5)
Partition
q:o
.
[ r
the resulting
coefficient
matrix
as
:= H,
Stability
(9)
rows and L( T + 1 + 0)
.!. .
4 1
(8)
Go to step (3).
conditions
H, is non-singular.
P=H,-[H_I...
HO_,].
expressing
X Ii&13
G. Anderson,
250
(IO)
4
Construct
:=
p:* .
G. Moore
Solwng
linecrr perfect
foresrght
models
the matrix
Concatenate
Q:=
.;.
[
(13)
initial
conditions
conditions,
I
.
Partition
[Q&Al
the auxiliary
Q,
:= Qv
cc>
X,:X,=
B,X,+;,t>,O
,= --7
is the unique
solution
converging
conditions.
4. An example
Suppose that the market rate of return is r, and consider the value of a declining firm whose
dividends are decaying to zero at rate S. The equation forecasting the value of the firm ex-dividend is
V,+1 = (1 + r)Y
while dividends
(6)
- D,+19
evolve according
D, = (1 - S)D,+,,
wherer>OandO<S<l.
to
(7)
G. Amkrson,
x,=
G. Moore
Solving
linear
perfect
foresight
The procedure
produces
(1-U
(r+ 6)
B -1=
i 0
(1-S)
As a check, recursively
substitute
+_~A_
Iterating
,+k
(9)
(1 - G)k+'D,_l.
Combining
I/=
value formula,
(8)
251
models
(1-aJ2D
(r+-6)
(10)
(-l
form computed
by the procedure.
5. Concluding remarks
The uniqueness and stability assumptions
exclude pathological models from consideration.
In the
Harrod-Domar
model [Solow (1969)] and Morishimas
(1973) model of Marxian theory, stable
solutions exist for only a subset of feasible initial conditions.
Other models such as Taylors (1977)
leading indicator
model and models described
in McCallum
(1981) have multiple
convergent
solutions for any initial conditions.
Our procedure indicates that the Harrod-Domar
model and
Morishimas
model have too many auxiliary initial conditions
and stability conditions
while the
leading indicator model has too few to determine a unique convergent solution.
A simple count of constraints is not sufficient to identify all pathological models. For example, the
model
q+1
= y,
z,=PZ,-
(11)7(12)
has solutions satisfying assumption (b) only with (~,/3 < 1. The model has the appropriate
number of
linear constraints when cy,/? > 1, even though no stable solution exists for any non-zero Z,_,. Simply
counting the number of eigenvalues on or outside the unit disc suggest the model might have a
reduced-form
solution, but checking the rank of QR indicates that LY,~> 1 is unacceptable.
This procedure for solving linear perfect foresight models serves as the nucleus of routines for
solving non-linear
perfect foresight models and for analyzing the covariance structure of solutions to
linear stochastic models. We discuss these topics in Anderson
and Moore (1984) and Moore and
Anderson (1984).
References
Anderson,
Gary and George Moore, 19R3, An efficient procedure
for solving linear perfect foresight models, unpublished
manuscript
(Board of Governors of the Federal Reserve System, Washington,
DC).
Anderson, Gary and George Moore, 19X4, An efficient procedure for solving nonlinear perfect foresight models, unpublished
manuscript
(Board of Governors of the Federal Reserve System, Washington,
DC).
McCallum.
Bennett, 1981, On uniqueness in rational expectations
models: An attempt at perspective,
Working paper 684
(NBER, Cambridge,
MA).
Moore. George and Gary Anderson, 1984, A state space method for solving linear rational expectations
models, unpublished
manuscript
(Board of Governors of the Federal Reserve System, Washington,
DC).
Morishima,
Michio, 1973. Marxs economics: A dual theory of value and growth (Cambridge
University Press. Cambridge).
Muth, John F.. 1961, Rational expectations
and the theory of price movements, Econometrica
29. 315-335.
Sargent. Thomas J., 1979, Macroeconomic
theory (Academic Press, New York).
Solow, Robert M., 1969, Growth theory (Oxford University Press, New York).
Stewart, G.W., 1976, Algorithm
506. HQR3 and EXCHNG:
FORTRAN
subroutines
for calculating
and ordering
the
eigenvalues of a real upper hessenburg matrix, ACM Transactions
on Mathematical
Software 2. 275-280.
Taylor, John, 1977, Conditions
for unique solutions
in stochastic
macroeconomic
models with rational
expectations,
Econometrica
45. 1377-1385.
Whiteman. Charles H., 1983, Linear rational expectations models: A users guide (University of Minnesota Press, Minneapolis,
MN).